Presenting Data Model Diagrams
... so people can understand them
Some principles/guidelines, with several examples, both good and bad.
TIME: 2+ hours, could be cut to 1 hour.
WHERE: DAMA EDW conference, Chicago, 2011; DAMA Minnesota Presentation, 2006 November.
First we must distinguish the notions of the semantics of a data model, from the syntax. Semantics is the underlying meaning of the model. The syntax of a data model refers to how it is written down, how it is depicted graphically. Syntax presents the semantics of the data model. Humans can only see and comprehend the semantics of a model, through the syntax, i.e., its presentation. Thus presentation becomes very important.
We often confuse syntax and semantics, as is evident in the criticisms of data models based on what we see in the presentation and not on the underlying semantics, which is the real model of interest.
Unlike the computer, humans have certain cognitive limitations. Presenting a data model with hundreds of entity types on large sheets of paper covering the wall of a conference room, is the antithesis of good presentation. We do not need to present the whole model all at once in all its detail. We can also be more creative than representing all entity types with a uniform "box." It is important for the user domain experts to understand a data model. After all, we are asking them to validate the correctness and completeness of the design.
This talk explores some principles for presenting data model diagrams in a way that people can more readily grasp the significant features in the model, and ultimately come to a full comprehension of the model semantics. These principles can be categorized as:
* Differentiation, layout - making the icons and their placement work for you
* Simplification: Horizontal abstraction - partitioning the model
* Simplification: Vertical abstraction - successive unfolding of detail.
* Focus - drawing the viewer's eye to what is most important.
* Navigation - how we can move around as we view the model; windowing.
These techniques are illustrated with several examples - both good and bad.
This talk was informed by a presentation given by and discussions with Dr. Kristin Potter, visualization researcher at the University of Oregon, Eugene, and by the doctoral research of Daniel Moody in Australia.